At expert level, the CRO's job is to run AI like any other revenue investment: with a portfolio, baselines, controlled tests and a clear view of cost and risk. Revenue operations is where that discipline lives.
The measurement stack
- Funnel baseline: volumes and conversion rates by stage and segment, before AI.
- Use-case register: each AI use case with owner, stage, target metric, cost and risk.
- Controlled tests: holdout groups, matched teams or staggered rollouts.
- Outcome metrics: pipeline, conversion, win rate, cycle time, quota attainment, gross and net retention.
- Cost and risk metrics: licences, oversight time, data work, complaints, unsubscribes, spam rates, errors found in reviews.
Measure the effect, not the score
A classic example from customer retention: in the Journal of Marketing Research (2018), Eva Ascarza showed that targeting customers for retention offers based on lift — how much the offer changes their behaviour — reduced churn more effectively than targeting those with the highest risk of leaving. A churn model that ranks risk perfectly can still waste the retention budget on customers who would have left (or stayed) anyway. The same logic applies to AI outreach and coaching: measure what changes because of the intervention.
Staggered rollouts
When a holdout feels unfair, roll out by region or team over several weeks. Teams that start later serve as the comparison for those that started earlier.
Report in revenue language
For each AI use case, report: the stage it targets, the measured change against comparison, the revenue that change is worth (use the funnel diagnostic), and the full cost. Retire use cases that don't pay back.
Sanity-check against the market
Consultancy and vendor research reports broad ranges for AI's revenue effect. Use them to check whether your plan is plausible — if your model says AI will double revenue, check your assumptions — but always decide on your own measured results.